Scanning transmission electron microscopy (STEM) thermometry techniques open new opportunities for mapping temperature (T) with high spatial resolution. Existing STEM thermometry methods based on measuring thermally-induced strains must contend with small thermal expansion coefficients (<10 parts per million (ppm)/K) for some materials, as well as potentially non-local relationships between strain and temperature. In contrast, the well-known mechanism of thermal diffuse scattering (TDS) offers promise for inherently local T measurements, and Debye-Waller theory predicts that many materials should display temperature coefficients >1000 ppm/K at room temperature. This T-dependent TDS has not been leveraged for STEM thermometry, however, because the Debye-Waller effect on the Bragg peak intensity is typically overwhelmed by the effects of thermal tilts and thermal drift.

Here, we demonstrate quantitative TDS measurements using STEM by measuring the diffuse background intensity (rather than the Bragg peak intensity) in energy-filtered scanning electron nanodiffraction patterns. Applying virtual apertures to these diffraction patterns during post-processing allows us to quantify the T-dependent TDS in the diffuse background between the Bragg spots; previous TEM work (with the beam in flood mode) showed that this diffuse signal is relatively insensitive to thermal tilts and drift. Using this diffuse signal, we measure a position-averaged temperature coefficient of 2400±400 ppm/K for a single-crystal gold film averaged between T=100 K and T=300 K, and compare our results with the predictions of Debye-Waller theory. The measurements display typical temperature uncertainties of 8 K and temperature sensitivities of 51 K Hz^-1/2. This TDS-based STEM thermometry demonstration provides a step towards the goal of non-contact nanoscale temperature mapping of thin nanostructures.